Modeling R&D Budget Profiles

Transcription

1 Modeling R&D Budget Profiles SCEA/ISPA Joint Annual Conference Orlando, FL June 2012 Erik Burgess

2 Background Agenda Key findings from 2004 put into practice Link between schedule and phasing Updated Models Space system phasing model Ground system phasing model Estimating with Variable Outlay Rates 2

3 Background 2004 National Intelligence Authorization Act Budgeting to the ICE becomes law Not just total, but every year Increased scrutiny on phasing models 2006 IC CAIG and NRO publish new models Four key enablers identified: 1. New accuracy metrics to defend model results 2. Improved regression methods for incorporating independent variables 3. New schedule models for defining start and end dates 4. Standard process for converting cost to budget Burgess, Erik. R&D Budget Profiles and Metrics. Journal of Parametrics, Volume XXV, Summer

4 Two Separate Models: Schedule and Phasing Schedule Estimating Relationship (SER) SER establishes nominal end date for phasing model Actual Duration Phasing Model NRO & DoD Data Annual $ Actuals Model Estimated Duration Time to First Launch = W.408 DL MT - 7.1OPT W = dry weight (lbs) DL = design life (months) MT = # of mission types (usually 1, e.g., comm) OPT = 1 if contract option Quality Metrics σ = 23% R 2 =.79 N = 82 Bias = 0% In practice, usually not enough money in early years, so what should we do? Decrease our cost estimate Slip schedule Argue for more funding 4

5 Our Most Powerful Accuracy Metric Cum-Cost Residual (% of total cost) Standard 40% complete Indicates confidence range through critical early years This model has σ = 9.8% Time In practice since 2006: Phasing model minus 1σ is minimum accepted funding request. Program schedules are slipped or funding is added. 5

6 Schedule and Phasing Are Linked Back- Loaded 50% 40% 38 NRO & DoD Programs 30% 20% Long Schedule 10% Long Schedule 0% -40% -30% -20% -10% 0% 10% 20% 30% 40% Among 18 back-loaded programs, only 1 beat our schedule model -10% -20% Among front-loaded programs, 76% beat our schedule model Back- Loaded -30% -40% -50% 6

7 Interpretation of These Data Any prediction that a contract will be completed with both 1. A back-loaded profile, and 2. A schedule faster than the CAIG baseline model is inconsistent with almost all historical data. Front-loading the budget is a necessary but not sufficient condition for programs to beat the CAIG schedule model. Other factors contribute to schedule delays. Scatter along the diagonal reflects error in the phasing model. Perfect phasing would fall on the diagonal due to error in schedule estimating. These data reflect final profiles and actual schedules, but contain no information on how programs were initially planned. 7

8 Satellite Expenditure-Phasing Model Weibull plus a constant-rate term 38 NRO & DoD Programs 387 time-cost pooled data points E() t = d Rt + 1 e total cost d = -α R + 1-e 0 t 1.0 β α t R = duration (mos.) α = X driver β = Y driver i i i i Relative Impact A α is a function of 4 drivers GFE % Sub BY07$M Duration Driver Coefficient (X) GFE (1,0) 1.84E+00 % Subs 2.73E-02 BY07$M 9.57E-04 Duration (mos) 2.79E-02 Driver Coefficient (Y) Competitive (1,0) 1.71E-01 GFE (1,0) 3.62E-01 % Subs 4.47E-03 BY07$M 7.03E-05 Duration (mos) -1.62E-03 Relative Impact B β is a function of 5 drivers COMP GFE % Sub BY07$M Duration 8

9 Ground Expenditure-Phasing Model Weibull plus a constant-rate term 28 IC & DoD Programs 224 time-cost pooled data points β α t ( ) E() t = d Rt + 1 e cost at t = 1. 0 d = -α R + 1-e α = β = Infrastructure or Terminal Follow-on ( ) R =.0011 Total Cost, BY09$M ACT RATE EST RATE Gives Range of Profiles Based on Independent Variables New Infrastructure or Terminal Front-Loaded 57/50 % % time Back-Loaded 35/50 Follow-on Data Processing, C2, or Mission Management 9

10 Expenditures Budget Authority estimates contract costs Final costs based on actual end-of-program historical data Annual expenditures based on actual expenditure profiles from completed programs Estimated expenditure profile is not a budget profile Budget authority must account for total government liability Difference between budget authority and expenditures is the annual outlay rate and others using published appropriation-wide outlay rates to convert expenditure estimate to budget request Process published by Lee, Hogue, and Gallagher in 1997 Implemented in our models since 2004 Lee, David A., Hogue, Michael R., and Gallagher, Mark A. Determining a Budget Profile from a R&D Cost Estimate, Journal of Cost Analysis,

11 Examples Large Development Contract Small Acquisition Contract Budget Expenditures Budget Authority Obligations Expenditures TY$M TY$M Fiscal Year Fiscal Year Budget Authority exceeds expenditures in early program years Several underlying causes not just poor performance Budgets often appear too front-loaded 11

12 What is Budget Authority Used For? A recently completed satellite contract 30-day carry per policy Open Commits Fee Budget Authority Required Budget Authority peaks in year 4 Subs & IWT Material & ODC Labor Accrued expenditures $5.4M per FTE needed in first year! Contract Year Labor peaks in year 6 Only14% of Ramp-up Budget Authority is In-house Labor Costs 12

13 Estimating Outlay Rates Outlay rates: Link between expenditures and Budget ( L ) BA s BA s BA s BA s k= ε k 2 k 1 3 k 2 J k J+ 1 1 Outlay rates, s i, have a large impact on budget in early years Appropriation-wide averages may not be appropriate Actual outlay rates vary during life of contract Program structures vary CAIG study approach: Collect data via CFSRs Actual government liability and expenditures each year Compare across contracts, over time, etc. Approach neutralizes effects of excessive or inadequate budget authority. 13

14 Basis for Analysis Example: Actual first year of Example contract. CFSR through September. Year 1 12) a) Open Commitments (CUM) 26,720 b) Accrued Expenditures (CUM) 23,149 c) Fee (CUM) 2,836 d) Total (CUM) (12a+12b+12c) 52,705 13) Estimated Termination Cost 1,100 14) Total Govt Liability (12d+13) 53,805 Incremental w/o Term Liability 15,439 15) Forecast of Billings to Govt (CUM) 23,169 $23,169 / $53,805 = 43% of liability was billed Exact budget has TOA matching line 14. True budget must have been greater or equal to liability. This is the actual amount billed that year In this example, exact budget would have a 43% year-1 outlay rate. Actual TOA cannot be lower than liability (by law). Actual outlay couldn t have been higher than 43%. 14

15 NRO Funding Policy CBP-20, 30 June 2010 Request obligation authority for additional 1 month of budget authority (carry forward) CFSR Year 1 12) a) Open Commitments (CUM) 26,720 b) Accrued Expenditures (CUM) 23,149 c) Fee (CUM) 2,836 d) Total (CUM) (12a+12b+12c) 52,705 13) Estimated Termination Cost 1,100 14) Total Govt Liability (12d+13) 53,805 Incremental w/o Term Liability 15,439 15) Forecast of Billings to Govt (CUM) 23,169 27,934 32,199 3,939 64,072 1,100 65,172 11,367 23,169 Additional 1 month of projected liability Same end-of year billing $23,169 / $65,172 = 36% of liability was billed In this example, realistic budget would have a 36% year-1 outlay rate. Actual budget may have been higher 15

16 Result for One Contract Realistic outlay rate computed each year Assume oldest money expended first 100% 90% 80% 70% Realistic Outlay Rate 60% 50% 40% 30% 20% 10% 0% Fiscal Year These are Realistic Annual Outlay Rates Assuming an Exact Budget Consistent with goal for Agency Cost Position & NRO Policy Consistent with actual program execution 16

17 Multiple Programs vs. Time Realistic Current Year Outlay vs Time 100% 90% Realistic Current-Year Outlay Rate 80% 70% 60% 50% 40% 30% 20% 10% 0% Today's practice = 58% every year (DoD-wide average, all contracts, all years) Contract Year Outlay rates increase gradually over the life of a contract Less open commitments and termination liability Less overall funds needed in future periods Difference among programs is highest in first few years 17

18 Application to Program Estimates We must predict outlay rates to use in estimates Two programmatic factors affect first-year outlay rate: 1 Month of ATP during the fiscal year Realistic Current-Year Outlay Rate 80% 70% 60% 50% 40% 30% 20% 10% 0% Year 1 Outlay Rate vs. Month of ATP Month of Contract ATP Realistic Current-Year Outlay Rate 80% 70% 60% 50% 40% 30% 20% 10% 2 Open commitments Year 1 Outlay Rate vs. Open Commitments Less open commitments means higher outlay rates 0% 0% 10% 20% 30% 40% 50% 60% Open Commitments (% of total liability) Note: These factors are correlated at

19 (1) Months Since ATP Affects later years as well Increasing trend can be modeled as a continuous function Implemented in space-segment phasing tool Realistic Current Year Outlay Rate vs. Months since ATP 100% 90% 80% Realistic Current-Year Outlay Rate 70% 60% 50% 40% 30% 20% 10% 0% Months since ATP 19

20 (2) Open Commitments Vs. Time 60% Open Commitments vs Time End of Fiscal Year Open Commitment % of Cum Liability 50% 40% 30% 20% 10% 0% 25% 13% 7% 3% 2% 1% 1% 0% 0% 0% 0% Contract Year Open commitments can be a high percentage of total liability in early years. At end of contract, vendors are delivering products, subcontracts contracts are closing out, new commitments are slowing. 20

21 Open Commits Drive Outlay Rates 100% 90% Year-1 Bllings/Liability (Max outlay rate) 80% 70% 60% 50% 40% 30% Less open commitments means higher outlay rates 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% Open Commitments (% of total liability) High level of open commitments drives outlay rates down. 21

22 What Drives Open Commitments? YR-1 Open Commitments (% of liability) 60% Open Commitments do not correlate with material 50% 40% 30% 20% 10% 0% 0.0% 2.0% 4.0% 6.0% 8.0% 10.0% 12.0% 14.0% 16.0% Percent Material (At End Of Contract) YR-1 Open Commitments (% of liability) 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% Open Commitments do not correlate with percent subcontracted Percent Subcontracted YR-1 Open Commitments (% of liability) 60% 50% 40% 30% 20% 10% Open Commitments do not correlate with spend rate Open commitments are driven by other factors Subcontract funding terms Accounting practices Predicting year-1 open commitments is difficult. 0% 0% 5% 10% 15% 20% 25% Year-1 Spend Rate (TGL % of EAC) Same result for other years. 22

23 CAIG s New Estimating Practice Avoid fixed outlay rates Use rates that increase during contract life. Allow tailoring to account for low or high open commitments. Realistic Current Year Outlay Rate vs. Months since ATP 100% 90% y = x R² = % Implement this curve in phasing tool. Allow tailoring points above curve have low open commits. Realistic Current-Year Outlay Rate 70% 60% 50% 40% 30% 20% 10% 0% Regression of historical data Months since ATP Today's practice = 58% every year (DoD-wide average, all contracts, all years) 23

24 Impact on Estimates: Example Underlying Expenditures Published Rates - same every year New Model - gradually increasing TY$M Contract Year Profile is less peaked 24

25 Allow users to input outlay rates by year Default to regression of historical data New Phasing Tool (Available to Industry) Adjusts weighted index for accurate BY-TY conversion New section computes rates based on time since ATP. Allows tailoring. 25

26

27 Definition of Terms Actual Max Outlay Rate cumulative forecast of billings to the government divided by the total government liability. This rate demonstrates the maximum percentage of budget a program manager could spend in a given period, assuming access to a perfect cost estimate. Realistic Outlay Rate calculated similarly to the max outlay rate except forecast of billings and total liability information estimated one month from current period. Open Commitments payment obligations legally binding the government to make payment in a given period. Accrued Expenditures authorized charges against available funds. Estimated Termination Cost the cost to the government of terminating a program prior to fulfillment of terms by the contractor. Forecast of Billings to Government expected amount to be invoiced to the government in a given period. Percent Subcontracted generally calculated here as total burdened subcontractor cost divided by total cost through G&A, when such program data is available. 27

28 Interpretation of α, β High ALPHA Low ALPHA ALPHA: Moves peak forward/backward Late Peak High Peak High BETA Slow ramp-up Slow ramp-up Early peak/front-loaded Low BETA BETA: Drives initial slope Fast ramp-up Fast ramp-up 28